Altered Protease–Activated Receptor-1 Expression and Signaling in a Malignant Pleural Mesothelioma Cell Line, NCI-H28, with Homozygous Deletion of the β-Catenin Gene

Protease activated receptors (PARs) are G-protein coupled receptors that are activated by an unique proteolytic mechanism. These receptors play crucial roles in hemostasis and thrombosis but also in inflammation and vascular development. PARs have also been implicated in tumor progression, invasion and metastasis. In this study, we investigated expression and signaling of PAR1 in nonmalignant pleural mesothelial (Met-5A) and malignant pleural mesothelioma (NCI-H28) cells. We found that the expression level of PAR1 was markedly higher in NCI-H28 cells compared to Met-5A and human primary mesothelial cells. Other three malignant pleural mesothelioma cell lines, i.e. REN, Ist-Mes2, and Mero-14, did not show any significant PAR1 over-expression compared to Met-5A cell line. Thrombin and PAR1 activating peptides enhanced Met-5A and NCI-H28 cell proliferation but in NCI-H28 cells higher thrombin concentrations were required to obtain the same proliferation increase. Similarly, thrombin caused extracellular signal-regulated kinase 1/2 activation in both cell lines but NCI-H28 cells responded at higher agonist concentrations. We also determined that PAR1 signaling through Gq and G12/13 proteins is severely altered in NCI-H28 cells compared to Met-5A cells. On the contrary, PAR1 signaling through Gi proteins was persistently maintained in NCI-H28 cells. Furthermore, we demonstrated a reduction of cell surface PAR1 expression in NCI-H28 and malignant pleural mesothelioma REN cells. Thus, our results provide evidences for dysfunctional PAR1 signaling in NCI-H28 cells together with reduced plasma membrane localization. The role of PAR1 in mesothelioma progression is just emerging and our observations can promote further investigations focused on this G-protein coupled receptor.


Introduction
Malignant mesothelioma (MM) is a relatively rare but highly aggressive neoplasm arising from mesothelial cells on the serosal surfaces of the pleural, peritoneal and pericardial cavities. Asbestos fiber exposure is widely accepted as the main cause with approximately 80% of cases being directly attributed to occupational exposure [1]. Although asbestos exposure has a pivotal role in initiating both cellular and molecular events which lead to MM development other factors such as genetic and epigenetic alterations contribute to its pathogenesis [1]. Several growth factors and their target receptors have been implicated in the oncogenesis, progression and resistance to therapy of MM [1]. In addition, the chemokine CXL12 and its target receptor CXCR4 which belongs to the large family of seven-transmembrane Gprotein coupled receptors (GPCRs), have been found to be highly expressed in malignant pleural mesothelioma (MPM) cell lines and tumor tissues suggesting they can be involved in tumor progression and survival [2].
Numerous evidences link aberrant GPCR expression and activation to several types of human malignancies [3,4]. Among GPCRs, PARs are a subset which have a unique mechanism of activation. In fact, they are activated enzymatically through proteolysis by enzymes of the serine protease family [5]. The proteolytic cleavage occurs at specific sites within their N-terminal region, thereby exposing novel N-termini, and the 'tethered ligand' then folds back onto the extracellular loop II of the receptor, resulting in activation. There are four PARs encoded by distinct genes in the mammalian genome. The prototype of this GPCR subfamily is PAR 1 which transmits cellular response to thrombin [6,7]. The receptor subfamily also includes PAR 2 which is activated by trypsin, and two other thrombin-activated receptors, PAR 3 and PAR 4 [8][9][10]. Other proteases besides trypsin for PAR 2 and thrombin and trypsin for PAR 1 and PAR 4 can activate these receptors [11]. Additionally, synthetic peptides that mimic the first six amino acids of the newly formed Nterminus can act as soluble ligands in the absence of receptor proteolysis. Activated PAR 1 couples to multiple heterotrimeric Gprotein subtypes including G i , G q and G 12/13 [11,12].
PARs have multiple roles in many physiological and pathological events involving different tissues and organs such as the cardiovascular, musculoskeletal, gastrointestinal, respiratory and central nervous system [13]. Coagulant proteases and PARs have been implicated in several types of malignant cancer. PAR 1 is over-expressed in aggressive melanoma, colon cancer, prostate cancer, and invasive breast cancer [14][15][16][17], promoting tumor cell invasion and epithelial cell malignancy [17][18][19][20]. In addition, several proteases, which can activate PAR 1 have been identified in tumors including tissue-derived trypsins, members of the coagulation cascade and matrix metalloprotease-1 [13,21]. Finally, a recent study have shown that MPM cell lines that express tissue factor and PAR 1 but not PAR 2 are able to generate large tumors in nude mouse throracic cavities [22].
In the present study, we analyzed PAR 1 expression levels, signaling and mitogenic effects in immortalized nonmalignant pleural mesothelial (Met-5A) and MPM cells (NCI-H28). In this MPM cell line, a homozygous deletion of the b-catenin gene (CTNNB1) has been demonstrated while thrombomodulin, a natural anticoagulant, appears to be silenced by an epigenetic mechanism [23,24]. Therefore, we were interested to study PAR 1 expression and signaling in this cell line and correlate our findings to known genetic and epigenetic alterations. Our work indicates that the expression levels of both PAR 1 mRNA and protein are increased in NCI-H28 cells compared to those found in Met-5A and primary human mesothelial cells. In addition, the increased PAR 1 expression appears to be an unique feature of the NCI-H28 cell line since in other three MPM cell lines, i.e. REN, Mero-14 and Ist-Mes2, PAR 1 levels are not significantly different from that found in Met-5A cells. Perhaps more important, PAR 1 signaling to down-stream effectors is dysfunctional as the signaling pathway through G i is the only one that is fully maintained while G 12/13 and G q pathways are reduced. Furthermore, the mitogenic effect triggered by PAR 1 activation is modified in NCI-H28 cells as compared to Met-5A cells. We also show that in this MPM cell line, cell surface PAR 1 expression is reduced and the receptor mainly localizes in the intracellular compartment. The intracellular retention of PAR 1 is likely responsible of the altered signaling.

Real time RT-PCR
RNA was isolated using the RNeasy Mini Kit (Qiagen) and tested for integrity by gel electrophoresis. mRNA was reverse transcribed to cDNA using a specific Rev Transcription Kit (New England BioLabs). Real time SYBR Green polymerase chain reaction (PCR) for PAR 1 was performed using forward primer: 59-TGCTTCAGTCTGTGCGG-39; and reverse primer: 59-CTCCATCAATAAAAGCAGTCCTCT-39. The relative expression of PAR 1 , with b-actin as the reference gene, was determined using the MiniOpticon Real-Time PCR Detection System (BioRad Laboratories, Inc., Hercules, CA, USA). Data are presented as expression ratios normalized to b-actin.
Immunocytochemistry NCI-H28 and Met-5A cells were seeded at 3610 4 cells per well in chamber slide (BD Biosciences, San Jose, CA, USA). Twentyfour hours later, cells were fixed in 2% paraformaldheyde in 0.1 M phosphate buffer, washed three times with PBS, rinsed, and blocked for 45 min with PBS containing 0.1% Triton-X 100 and 1% BSA. After washing, cells were incubated with mouse monoclonal anti-PAR 1 (1:100), mouse monoclonal anti-b-catenin (1:500) or rabbit polyclonal anti-b-catenin (1:1000) and rabbit polyclonal anti-caveolin-1 (1:400) primary antibodies diluted in PBS containing 0.03% Triton-X 100 and 1% BSA for 18 h at 4uC. Double labelling studies were carried out as follow: anti-PAR 1 and anti-caveolin-1; anti-PAR 1 and rabbit polyclonal antib-catenin; mouse monoclonal anti-b-catenin and anti-caveolin-1. After washing, to visualize single staining, cells were incubated with Alexa Fluor 488-and Alexa Fluor 568-labeled goat antimouse (1:400) or anti-rabbit (1:400) antibodies for 2 hour at room temperature. Then slides were covered with Vectashield (Vector Laboratories, Burlingame, CA, USA). Confocal images were obtained with a Leica TCS-SP5 confocal microscope, using a 406oil objective with 1.45 NA and a recommended pinhole size of less than 1.0 micrometer. The images were processed with PhotoshopCS3 software. To evaluate fluorescence colocalization, the images were also analyzed using the freely available ImageJ program [34].

Cell proliferation assay
Met-5A and NCI-H28 cells were plated at 3610 3 cells/well in clear 96-well dishes and allowed to adhere overnight. Then cells were serum and growth factor starved for 12 hours and stimulated with and without agonists for 72 hours. After that, 10 ml of WST-1 mixture was added to each well, mixed gently for one min and cells incubated for additionally 2 hours at 37uC. Finally, the formazan dye was quantified by measuring the absorbance of each sample against background as blank with a Wallac 1420 multilabel counter microplate reader (PerkinElmer, Inc., Boston, MA, USA) at a wavelength of 450 nm.
[Ca 2+ ] i measurement PAR-induced increase of [Ca 2+ ] i was assessed by measuring fluorescence variations after agonist stimulation of cells loaded with Fluo-3 AM using a Wallac 1420 multilabel counter microplate reader (PerkinElmer, Inc., Boston, MA, USA), as previously described [32]. Cells were seeded in black/clear bottom 96-well assay plates at a density of 2610 4 cells/well (HMEC-1) or 1.5610 4 cells/well (Met-5A and NCI-H28) in complete growth media. After attachment, cells were starved in serum and growth factor free media containing BSA for 3 h at 37uC. Before starting the assay, cells were washed twice with loading buffer (20 mM Hepes, 0.83 mM Na 2 HPO 4 , 0.17 mM NaH 2 PO4, pH 7.4, 130 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , and 1 mM MgSO 4 ) containing 25 mM mannose, 1 mg/ml BSA and 2.5 mM probenecid and then incubated in 100 ml of the same buffer containing 6 mM Fluo-3 AM/0.024% pluronic acid. After 1 h at 37uC, cells were washed twice with loading buffer and incubated in 100 ml of the same buffer for an additional 1 h at 37uC. Fluorescence was recorded at baseline and every 3 seconds after thrombin (10 nM) or PAR 1 -APs (10 mM) addition for another 120 seconds.

RhoA activation assay
Levels of GTP-bound RhoA were determined in serum and growth factor starved (18 h) Met-5A and NCI-H28 cells before and 2 min after stimulation with 10 nM thrombin or 10 mM selective PAR 1 -AP using a G-LISA RhoA activation assay kit (Cytoskeleton, Denver, CO, USA).

Measurement of intracellular cAMP
Intracellular cAMP levels were measured using a competitive protein binding method as previously described [32]. Met-5A and NCI-H28 cells (4610 4 /well) were plated in 24-well dishes and allowed to grow for 24 h. Thereafter, cells were incubated for 15 min in serum and growth factor free media containing 20 mM 4-(3-Butoxy-4-methoxybenzyl)-2-imidazolidinone and then exposed to different thrombin or selective PAR 1 -AP concentrations in the presence and absence of 100 nM SCH 79797 for 15 min. Assays were initiated by the addition of 1 mM isoproterenol.

Cell surface ELISA
Detection of endogenous PAR 1 expressed on the cell surface was quantified by ELISA essentially as described by Paing et al. [35]. Met-5A and NCI-H28 cells were plated in 24-well dishes at 6610 4 cell/well and grown overnight. Assay media was RPMI-1640 supplemented with 1 mg/ml BSA and 1% penicillin/ streptomycin. Cells were washed with media and incubated on ice for 30 min. Afterwards, cells were washed and incubated with 5 mM SCH 79797 for 30 min and then treated with 10 nM thrombin for 10 min at 37uC. Cells were then washed with PBS and fixed with 4% paraformaldehyde on ice for 10 min. After fixation, cells were washed with PBS and incubated with primary antobody for 1 h at room temperature, followed by incubation with horseradish peroxidase-conjugate goat anti-rabbit secondary antibody for 1 h at room temperature. After washing, cells were incubated with horseradish peroxidase substrate 1 step 2,29azinobis (3-ethylbenzthiazoline-6-sulfonic acid) diammonium salt at room temperature. An aliquot was removed from each well and optical density was determined at 405 nm using a Wallac 1420 multilabel counter microplate reader.
Transient b-catenin transfection and RNA interference NCI-H28 and Met-5A cells were seeded onto 24-well plates at 5610 4 cells/well and transfected 24 h later with 0.7 mg/well pCMV6XL5-b-catenin or empty vector and 30 nM b-catenin or scrambled non-targeting siRNAs, respectively. Transfections were carried out for 48 h in RPMI-1640 using Lipofectamine 2000 according to the manufacturer's suggested conditions (Invitrogen). ELISA assays for detection of cell surface PAR 1 in transfected cells were performed as described above.

Results
PARs and their potential activating proteases are frequently over-expressed in human tumor tissues, including prostate cancer, invasive breast cancer, colon cancer, and malignant melanoma [14][15][16][17][18][19][20]. Lee et al. [36] have shown that PAR 2 is present in human pleural tissues where it plays a role in pleural inflammatory responses while in primary cultures of human peritoneal mesothelial cells the expression of PAR 1 has been reported [37]. In addition, the expression of PAR 1 has been revealed in 3 MPM cell lines by western blot analysis but these cell lines do not express PAR 2 [22]. Therefore, we decided to investigate expression and signaling of PAR 1 in human pleural mesothelial and MPM cells to evaluate the possible role of this receptor in mesothelioma cell proliferation. For this work we utilized the MPM cell line, NCI-H28, which does not express CXCR4 and the nonmalignant pleural mesothelial cell line, Met-5A, was used as a control [2]. In this MPM cell line, apart from a homozygous deletion of the bcatenin gene (CTNNB1) a down-regulation of thrombomodulin expression by an epigenetic mechanism has been described [23,24]. The expression of thrombomodulin, a glycosylated transmembrane protein which binds with high affinity to thrombin inhibiting its enzymatic activity and accelerating protein C activation, is lower in MPM tissue than in normal mesothelium [24]. In addition, low or no expression of thrombomodulin in various cancers has been associated with poor prognosis [38][39][40].

PAR 1 is over-expressed in NCI-H28 cells
To verify whether PAR 1 mRNA level was different in malignant NCI-H28 cells compared to nonmalignant Met-5A cells, real time RT-PCR was performed using RNA extracted from these cells. In NCI-H28 cells, PAR 1 mRNA level was significantly increased compared to Met-5A cells (Figure 1.A). Immunoblot analysis showed a 48 kDa band corresponding to PAR 1 in lysates of Met-5A, NCI-H28 and other three MPM (Ist-Mes2, REN and Mero-14) cell lines while two close bands were detectable in immunoblot of human primary mesothelial cell lysates (Figure 1.C). The appearance of two bands was not a surprise since human PAR 1 contains multiple glycosylation consensus sites and several studies have shown the detection of 40 to 100 kDa bands on immunoblots [41][42][43]. However, the PAR 1 protein expression was lower in primary mesothelial cells than in Met-5A cells (Figure 1.B and 1.C). In NCI-H28 cells, the protein expression level was significantly increased compared to primary mesothelial and Met-5A cells (Figure 1.B and 1.C). In the other MPM cell lines, PAR 1 protein levels were essentially similar to that found in Met-5A cells. Therefore, the increased PAR 1 expression is an unique feature of NCI-H28 cell line. Overall, these findings suggest that the increased expression of PAR 1 in NCI-H28 cells results from increased gene transcription although enhanced PAR 1 mRNA and/or PAR 1 protein stability can also be involved. We also examined PAR 2 mRNA and protein levels in Met-5A and NCI-H28 cells (data not shown). Real time RT-PCR and western blot analysis demonstrated PAR 2 expression levels were similar in both cell lines. Fluorescence was recorded before agonist addition (F 0 ) and then every 3 seconds after thrombin (10 nM) or PAR 1 -AP (10 mM) addition for another 120 seconds. Data shown are mean 6 SEM of a single experiment done in triplicate. Experiments were repeated two additional times with similar results. The results are reported as relative fluorescence (RF = F/F 0 where F 0 is basal fluorescence and F is fluorescence recorded after cell stimulation with the agonist). C, inhibition of isoproterenol stimulated cAMP production in Met-5A and NCI-H28 cells by different concentrations of thrombin in the presence and absence of 100 nM SCH 79797. D, no effect of the selective PAR 1 -AP on isoproterenol stimulated cAMP production in Met-5A and NCI-H28 cells. Serum and growth factor starved cells were exposed to different agonist concentrations. Assays were initiated by the addition of 1 mM isoproterenol. Production of cAMP was measured using a competition binding assay which includes the bovine adrenal cAMP binding protein and [ 3 H]cAMP. Data shown are mean 6 SEM of three independent experiments performed in triplicate. The differences between thrombin-and thrombin plus SCH 79797-treated cells were significant (**P#0.01, ***P#0.001) by one-way ANOVA followed by Bonferroni's multiple comparison test (n = 3). doi:10.1371/journal.pone.0111550.g004 PAR 1 agonists enhance Met-5A and NCI-H28 cell proliferation Next, we examined whether in NCI-H28 cells, PAR 1 was functionally active by evaluating thrombin-or PAR 1 -APs-induced cell proliferation. Met-5A and NCI-H28 cells were incubated with various thrombin or PAR 1 -AP concentrations for 72 h. In Figure 2.A, the proliferative responses induced by thrombin stimulation are reported. Both Met-5A and NCI-H28 cells showed significant increases of cell proliferation at 72 h (Figure 2.A). However, the pattern of the proliferative response was quite different in NCI-H28 cells compared to that of Met-5A cells. As an example, in Met-5A the proliferative response was maximal at 1 nM thrombin with a progressive decrease up to 50 nM while in NCI-H28 cells the maximal response was reached at 50 nM (Figure 2.A).
The non-selective PAR 1 -AP, SFLLRN-NH 2 , was less effective than thrombin in stimulating Met-5A and NCI-H28 cell proliferation (Figure 2.B). A 24-28% increase of cell proliferation was reached at 10 and 100 mM SFLLRN-NH 2 in Met-5A and NCI-H28 cells, respectively (Figure 2.B). The selective PAR 1 -AP, Figure 5. PAR 1 agonist-induced G 12/13 signaling is impaired in NCI-H28 cells. A, relative levels of RhoA activation in response to thrombin in Met-5A and NCI-H28 cells. B, relative levels of RhoA activation in response to the selective PAR 1 -AP in Met-5A and NCI-H28 cells. Rho A activation was measured in serum and growth factor starved cells using the RhoA G-LISA kit from Cytoskeleton. Data shown are mean 6 SEM of three independent experiments performed in triplicate. The differences in RhoA activation between Ctrl (vehicle treated Met-5A or NCI-H28 cells) and agonist-treated cells were significant (*P#0.05, **P#0.01, ***P#0.001) by one-way ANOVA followed by Bonferroni's multiple comparison test (n = 3). doi:10.1371/journal.pone.0111550.g005 Altered PAR1 Signaling in a Mesothelioma Cell Line TFLLR-NH 2 , was less efficacious in stimulating cell proliferation than SFLLRN-NH 2 but a concentration of 100 mM caused a 20% increase of NCI-H28 cell proliferation (Figure 2.C). These results highlight that PAR 1 -APs do not behave exactly as thrombin in stimulating cell proliferation.

Reduced cell surface PAR 1 expression in NCI-H28 cells
Since NCI-H28 cells respond with proliferation at higher thrombin concentrations even though they express increased PAR 1 levels (see Figure 1), we questioned whether the receptor is properly localized on cell surface in this cell line. Therefore, we compared the amount of cell surface PAR 1 in Met-5A, NCI-H28 and REN cells using an ELISA assay. Interestingly, NCI-H28 cells showed significantly less cell surface PAR 1 expression than Met-5A cells (Figure 3). REN cells, which express b-catenin as indicated by immunoblot analysis (data not shown), also showed a reduced cell surface receptor expression compared to Met-5A cells (Figure 3). Overall, these findings provide evidences of an altered cell surface distribution of PAR 1 in two MPM cells lines showing different levels of total receptor expression.

Dysfunctional PAR 1 signaling in NCI-H28 cells
To further explore PAR 1 ability of signaling in the NCI-H28 cell line, receptor-activated G q , G i , and G 12/13 signaling pathways Figure 6. Thrombin differently induces ERK1/2 activation in Met-5A and NCI-H28 cells. A, relative intensity of pERK1/2 immunoreactive bands quantified by densitometric scanning. Serum and growth factor starved Met-5A and NCI-H28 cells were incubated in the presence and absence of various thrombin concentrations ranging from 0.01 to 100 nM for 5 min. ERK1/2 activation was then determined using a specific anti-phospho-ERK1/2 antibody. Nitrocellulose membranes were then stripped and reprobed for total ERK1/2. Data (mean 6 SEM) are expressed as fold-increase over Ctrl and are the averages of three independent experiments performed in duplicate. The differences in phosphorylated ERK1/2 level between Ctrl (vehicle treated Met-5A or NCI-H28 cells) and thrombin-treated cells were significant (*P#0.05, **P#0.01) by one-way ANOVA followed by Bonferroni's multiple comparison test. B, a representative immunoblot. doi:10.1371/journal.pone.0111550.g006 were examined. First, we investigated PAR 1 -activated G q signaling by analyzing intracellular Ca 2+ mobilization after cell stimulation with either thrombin or the selective PAR 1 -AP. As indicated by relative fluorescence increase, both thrombin (10 nM) and PAR 1    components of the G q signaling pathway by immunoblot analysis ( Figure S1). Whereas PLC-b 1 was expressed at similar levels in both cell lines, the amount of Ga q was apparently greater in NCI-H28 than Met-5A cells ( Figure S1).
To explore the functional integrity of G i signaling pathway, we analyzed thrombin-and PAR 1 -AP-induced inhibition of isoproterenol stimulated cAMP accumulation in both Met-5A and NCI-H28 cells. In Met-5A cells, 10 pM to 1 nM thrombin inhibited isoproterenol stimulated cAMP production in a concentration dependent manner reaching 50% inhibition at 1 nM (Figure 4.C). However, at higher thrombin concentrations (1 nM to 100 nM) the inhibitory effect was progressively diminished. In the presence of SCH 79797, the inhibitory effect of thrombin was reduced indicating that PAR 1 mediates the effect. In NCI-H28 cells, thrombin inhibited cAMP in a concentration dependent manner reaching 50% and maximal inhibition (approximately 70%) at 1 nM and 100 nM, respectively (Figure 4.C). In the presence of SCH 79797, the inhibition curve was upwards shifted and the maximal inhibition at 100 nM was only 42% indicating that the inhibitory effect of cAMP accumulation is partially mediated by PAR 1 . Various concentrations of the selective PAR 1 -AP did not cause any inhibition of isoproterenol stimulated cAMP production in both Met-5A and NCI-H28 cells (Figure 4.D) demonstrating the functional selectivity of this peptide agonist.
Next, we examined PAR 1 -activated G 12/13 signaling by measuring RhoA activation after cell stimulation with either thrombin or PAR 1 -AP. In Met-5A cells, 10 nM thrombin induced a significant 2.5-fold increase of RhoA activation while in NCI-H28 cells the increase was just 1.2-fold ( Figure 5.A). The selective PAR 1 -AP (10 mM) was less effective in stimulating RhoA activation than thrombin in Met-5A cells but it still caused a significant increase (Figure 5.B). Similarly to thrombin, PAR 1 -AP induced a modest increase of RhoA activation in NCI-H28 cells ( Figure 5.B). We also examined the expression levels of Ga 12 , Ga 13 , and RhoA in both cell lines by immunoblot analysis ( Figure  S1). Our results indicate Ga 12 and RhoA expression levels were similar in Met-5A and NCI-H28 cells while Ga 13 expression was significantly increased in NCI-H28 cells compared to Met-5A cells ( Figure S1).
To further investigate distinctions in signaling, we examined thrombin induced ERK1/2 activation, an important mitogenic signaling cascade, in Met-5A and NCI-H28 cells. Thrombin (10 nM) caused a rapid increase of phosphorylated-ERK1/2 (pERK1/2) which reached a maximum level at 5 min and persisted up to 30 min in both cell lines (data not shown). Using a single time point (5 min) we examined the effect of various thrombin concentrations ranging from 0.01 to 100 nM and found that a maximal response was induced by 0.1 nM thrombin in Met-5A cells while higher thrombin concentrations reduced pERK1/2 Figure 9. Neither b-catenin rescue nor deletion modify cell surface PAR 1 expression. NCI-H28 cells were transiently transfected with plasmide vector containing CTNNB1 or empty vector (Ctrl) while Met-5A cells were transfected with nonspecific (Ctrls) or specific bcatenin siRNA as described in Materials and Methods. A, relative expression levels of b-catenin. Transfected cells were lysed and total cell proteins were analysed by immunoblot using an anti-b-catenin antibody. Then membranes were reprobed with an anti-b-actin antibody. Data are expressed as arbitrary unit (fold variation over Ctrl) after normalization by b-actin. Data shown are mean 6 SEM of three independent experiments. The differences of b-catenin relative levels between Ctrls and cell transfected with the recombinant vector or specific siRNA were significant (*P#0.05) by one-way ANOVA followed by Bonferroni's multiple comparison test (n = 3). B, a representative immunoblot. C, cell surface PAR 1 expression measured by ELISA assay. Antibody binding to fixed transfected cells was detected by horseradish peroxidise-conjugated secondary antibody. Data represent the mean 6 SEM of three independent experiments performed in triplicate. The differences in cell surface PAR 1 expression between Ctrls and cell transfected with the recombinant vector or specific siRNA were significant (***P#0.001) by one-way ANOVA followed by Bonferroni's multiple comparison test (n = 3). doi:10.1371/journal.pone.0111550.g009 ( Figure 6). In contrast, NCI-H28 cells demonstrated maximal pERK1/2 activity at 10 nM thrombin ( Figure 6). Of note, PAR 1induced ERK1/2 phosphorylation patterns in Met-5A and NCI-H28 cells were quite similar to respective thrombin-induced cell proliferation profiles (Figure 2.A).

Prevalent intracellular PAR 1 localization in NCI-H28 cells
In human umbilical vein endothelial cells, it has been reported that b-catenin greatly facilitates recruitment of caveolin-1 to VEcadherin/catenin complex at cell junctions [44]. Additionally, several lines of evidence indicate that caveolae are relevant for GPCRs/G proteins signaling including that driven by PAR 1 [45,46]. As NCI-H28 cells have a homozygous deletion of the bcatenin gene we questioned whether the lack of this protein could reduce cell membrane recruitment of both caveolin-1 and PAR 1 . Therefore, we analyzed b-catenin, caveolin-1 and PAR 1 localization in Met-5A and NCI-H28 cells by immunocytochemistry (Figure 7 and 8). In Met-5A cells, both b-catenin and caveolin-1 were localized on the plasma membrane including at some cell junctions and PAR 1 also showed a prevalent but not exclusive localization on the plasma membrane ( Figure 7). In contrast, in NCI-H28 cells there was no b-catenin staining, and caveolin-1 and PAR 1 were mainly localized in the cytoplasm (Figure 7). In Met-5A cells, double labeling studies suggested b-catenin and caveolin-1 closely localized at cell junctions. In addition, both intracellular and plasma membrane PAR 1 apparently colocalized with caveolin-1 ( Figure 8). In NCI-H28 cells, the intracellular PAR 1 was also in close proximity to caveolin-1 as suggested by the yellow stain ( Figure 8). A quantification of PAR 1 /caveolin-1 colocalization using Pearson's correlation coefficient (PCC) indicated a good degree of correlation in both Met-5A (PCC = 0.7760.05; n = 6) and NCI-H28 cells (PCC = 0.8460.03; n = 6).
Neither b-catenin rescue nor deletion affect cell surface PAR 1

expression
In order to test our hypothesis that b-catenin is required for proper cell surface PAR 1 localization, we transiently transfected NCI-H28 cells with a plasmide vector containing human b-catenin cDNA and silenced b-catenin expression in Met-5A cells using a specific siRNA. Immunoblot analysis indicated that in NCI-H28 cells transfected with the recombinant vector, b-catenin was expressed at high levels ( Figure 9.A and 9.B) compared to the expression level in cells transfected with the empty vector. On the other hand, we also obtained a consistent reduction of b-catenin expression in Met-5A cells transfected with the b-catenin siRNA as compared to cells treated with a nonspecific scrambled siRNA (Figure 9.A and 9.B). However, in ELISA assays b-catenin transfected NCI-H28 cells did not show any increase of cell surface PAR 1 expression while silenced Met-5A cells had no significant decrease of cell surface receptor as compared to control cells (Figure 9.C). Using immunofluorescence microscopy, we were also unable to detected any important change of PAR 1 localization in b-catenin transfected and silenced cells as compored to respective controls (data not shown). All together, our findings indicate that the lack of b-catenin is not responsible for reduced cell surface PAR 1 localization.

Discussion
Coagulant proteases and PARs have been implicated in several types of malignant tumors. Indeed, a well-documented link between hyperactivation of the coagulation cascade and tumor progression exists. The pro-coagulant activity mediated by the action of coagulant proteases such as thrombin can contribute to the malignant phenotype both directly, by stimulating tumor cell proliferation, and indirectly through the development of tumorassociated thromboemboli [47]. Among cancer patients, those with MPM are very susceptible to thromboembolic complications [48]. In addition, Keshava et al. [22] have shown that MPM cell lines, which express tissue factor and PAR 1 generate large tumors in mouse thoracic cavity thus indicating that activation of PAR 1 promotes MPM cell growth.
To this end we investigated whether a correlation exists between PAR 1 expression and cell proliferation using a MPM cell line (NCI-H28) and a nonmalignant pleural mesothelial cell line (Met-5A). In the NCI-H28 cell line, thrombomodulin, a transmembrane glycoprotein that controls thrombin-mediated proteolysis, is silenced by an epigenetic mechanism [24]. We found that the proliferative response of NCI-H28 cells to various thrombin concentrations was quite different from that obtained with the nonmalignant pleural mesothelial cell line. Whereas in NCI-H28 cells, thrombin-induced proliferation increased in a concentration dependent fashion, in Met-5A cells thrombin induced the maximal effect at 1 nM and then at higher concentrations the stimulatory effect progressively decreased. The proliferative response of NCI-H28 cells increased without reaching any growth steady state as expected when cells lose contact inhibition, a typical characteristic of cancer cells. The diverse response can result as consequence of reduced cell surface localization of PAR 1 (Figure 3) in NCI-H28 cells even though the total receptor amount is increased. However, we do not feel to exclude that the lack of thrombomodulin in NCI-H28 cells [24] affects PAR 1 growth signaling.
The non-selective PAR 1 -AP, SFLLRN-NH 2 , enhanced proliferation of both nonmalignant pleural mesothelial and MPM cells in a concentration-dependent fashion [13]. However, the proliferative response was slightly less marked than that observed with thrombin suggesting that either thrombin is also acting through other receptors or PAR 1 activation by proteolytic cleavage elicits a cellular response which is not completely identical to that induced by a ''free'' synthetic peptide agonist. Backhart et al. [49] have reported that distinct cellular responses can be evoked by thrombin versus synthetic peptide agonists. In addition, McLaughlin et al. [50] have demonstrated that thrombin-activated PAR 1 preferentially couples to G 12/13 proteins while PAR 1 -APs favor activation of G q signaling leading to [Ca 2+ ] i increase. The modest enhance of cell proliferation induced by the selective PAR 1 -AP suggests that PAR 2 may also contribute to thrombin-and SFLLRN-NH 2-stimulated functional response in both cell lines. Although thrombin is not able to cleave and activate PAR 2 , thrombin-cleaved PAR 1 can transactivate PAR 2 in human umbilical vein endothelial cells [51]. Indeed, as mentioned before, we were able to detect similar levels of PAR 2 expression in Met-5A and NCI-H28 cells (data not shown).
When PAR 1 -mediated activation of signaling pathways was examined, we immediately noticed that G q and G 12/13 signaling was compromised in NCI-H28 cells. In this MPM cell line, the only signaling pathway which was fully activated by thrombincleaved PAR 1 is through G i proteins leading to inhibition of adenylyl cyclase. Indeed, thrombin inhibited cAMP production in a concentration-dependent fashion in NCI-H28 cells while in Met-5A cells it showed a biphasic effect. Simultaneous activation of different G proteins with release of a plethora of Gbc subunits which are able to activate some isoforms of adenylyl cyclase can be responsible for the biphasic shape of the curve [52]. It is interesting to note that the selective PAR 1 -AP did not cause any major inhibition of cAMP accumulation. These findings are in agreement with thrombin and PAR 1 -AP displaying functional selectivity at PAR 1 as reported by McLaughlin et al. [50].
Decreased G q and G 12/13 signaling with the prevalence of G i signaling can explain the altered proliferative response to thrombin in NCI-H28 cells. Indeed, PAR 1 -mediated activation of ERK1/2 occurs through both G q and G i signaling with consequent activation of mitogenesis [53]. When we examined thrombin-induced ERK1/2 activation we found that lower thrombin concentrations were able to activate ERK1/2 in Met-5A than in NCI-H28 cells. This finding supports the role of G q signaling in mediating thrombin-induced ERK1/2 activation in Met-5A. Persistent PAR 1 signaling as consequence of altered receptor trafficking has been reported in metastatic breast carcinoma cells leading to enhanced cellular invasion [20]. We might speculate that altered PAR 1 signaling can also impact MPM cell invasiveness.
Compartmentalization of PARs and G proteins in plasma membrane lipid raft microdomains such as caveolae can confer PAR/G protein selectivity [21]. Russo et al. [46] have shown the critical role of caveole in activated protein C (APC) activation of PAR 1 selective signaling in endothelial cells. Furthermore, some studies concerning other GPCRs have demonstrated that caveolin-1 is required to prolong G q signaling and inhibit receptor coupling to G i/o proteins [54,55]. In thrombin-stimulated endothelial cells, caveolin-1 opens cell junction by targeting catenins [44]. The recruitment of caveolin-1 at cell junctions is greatly facilitated by the presence of b-catenin in the cadherin/catenin complex. In NCI-H28 cells, a homozygous deletion of the b-catenin gene (CTNNB1) has been demonstrated suggesting that in these cells caveolin-1 is not completely associated to the plasma membrane [21]. Our immune fluorescence experiments show that in NCI-H28 cells caveolin-1 is partially retained in the cytoplasm while in Met-5A cells it is prevalently localized to the plasma membrane. In Met-5A cells, PAR 1 is distributed in both plasma membrane and intracellular compartments and double immunolabeling studies suggest its proximity to caveolin-1. In NCI-H28 cells, PAR 1 is mostly retained in the intracellular compartment. Of note, PAR 1 and caveolin-1 appear to colocalize in both cell lines as suggested by PCC values. The intracellular retention of the receptor is confirmed by ELISA showing a consistent reduction of cell surface PAR 1 in NCI-H28 cells compared to Met-5A cells. However, we do not know whether in NCI-H28 cells the increased intracellular receptor distribution is due to altered cell surface recruitment or enhanced-internalization of activated receptor. Of note, REN cells, another MPM cell line, which express similar PAR 1 levels than Met-5A cells, also show a reduction of cell surface PAR 1 by ELISA assay (see Figure 3). This aggressive MPM cell line does not express thrombomodulin as the NCI-H28 cell line and expresses high levels of tissue factor and very little amount of endothelial cell protein C receptor [22]. Thus, these evidences suggest that the observed reduction of cell surface PAR 1 expression in these MPM cell lines can result as consequence of activated-receptor internalization. In order to exclude a role of bcatenin in recruiting PAR 1 to the plasma membrane, we performed both rescue and deletion experiments and evaluated cell surface receptor expression by ELISA. However, our findings indicate that b-catenin expression is not required for cell surface PAR 1 localization in both NCI-H28 and Met-5A cells. Since the NCI-H28 cell line is only one among other MPM cell lines examined, which shows a highly significant increase of PAR 1 expression compared to Met-5A and human primary mesothelial cells, we may speculate that b-catenin indirectly modulates PAR 1 expression at transcriptional level.
In summary, we have demonstrated that PAR 1 is highly overexpressed in a MPM cell line, NCI-H28, while other three MPM cell lines show similar or slightly increased expression levels than a mesothelial cell line (Met-5A) and human primary mesothelial cells. Thrombin promotes Met-5A and NCI-H28 cells proliferation through activation of PAR 1 . In NCI-H28 cells, PAR 1 although over-expressed, is defective in cell surface localization and signaling through G q and G 12/13 pathways. Cell surface PAR 1 expression is also reduced in MPM REN cells, thus suggesting receptor activation and internalization by cell produced proteases in both cell lines. Further studies are needed to investigate the role of cell surface or secreted proteases in inducing PAR 1 activation and stimulation of MPM growth. Figure S1 Expression of Ga subunits, RhoA, PLCb 1 , and caveolin-1 in Met-5A and NCI-H28 cells. Cells were lysed and protein solubilized as described under Materials and Methods. Proteins were then separated by SDS-PAGE and transferred onto nitrocellulose. Specific anti-Ga q , -Ga 12 , -G 13 , -RhoA, -PLCb 1 , and -caveolin-1 antibodies were used to detect each protein.

Supporting Information
Nitrocellulose membranes were subsequently stripped and reprobed with an anti-b-actin antibody. The intensity of the immunoreactive bands was quantified by densitometric scanning. A, relative intensity of immunoreactive bands. Data are expressed as arbitrary unit (fold increase over Ctrl, Met-5A) after normalization by b-actin. Data shown are mean 6 SEM of three independent experiments. The differences in protein expression between Met-5A and NCI-H28 cells were significant (*P#0.05) by one-way ANOVA followed by Bonferroni's multiple comparison test (n = 3). B, a representative immunoblot. Polyclonal anti-Ga antibodies were obtained from ABCAM (Cambridge, UK) while monoclonal anti-RhoA and polyclonal anti-PLCb 1 antibodies were from EMD Millipore Biosciences (Billerica, MA) and Thermo Fisher Scientific (Waltham, MA), respectively. (TIF)